The detection of lesions in conventional mammography is a difficult task, predominantly due to the masking effect of superimposed parenchyma! breast patterns. Limited angle, tomographic mammography, also referred to as breast tomosynthesis, is a technique that has been proposed to reduce this masking effect, by providing the radiologist with tomographic image slices through the breast. The goal of the proposed research is to investigate the use of iterative reconstruction methods for breast tomosynthesis. Iterative reconstruction methods have a number of potential advantages over some previously proposed tomosynthesis methods including;1) a more accurate modeling of the noise in the data, 2) the capability for modeling the physics of x-ray transport, thus providing an integrated approach for compensation of scatter and detector blur, and 3) the capability of incorporating a priori information on the object to be reconstructed. Our hypothesis is that breast tomosynthesis using iterative reconstruction methods can provide improved detection of malignant lesions as compared to backprojection tomosynthesis, as well as to conventional two-view digital mammography. To test this hypothesis, human observer psychophysical studies will be performed comparing conventional two-view digital mammography and tomosynthesis with different reconstruction approaches using patient data acquired with a commercial prototype breast tomosynthesis system. We also propose to investigate a number of issues related to the acquisition process of breast tomosynthesis including;1) alternative acquisition geometries, 2) the impact of varying levels of breast compression, 3) the impact of scatter, and 4) the optimal anti-scatter grid. Evaluation and optimization of different imaging system designs and acquisition processes will be conducted by evaluating lesion detection accuracy using realistically simulated tomosynthesis breast images.